A sparse and structured constraint-based method for large-scale economic data analysis
PeerJ Computer Science · 05.06.2026 · 12 · e3895
DOI: 10.7717/peerj-cs.3895 ↗Kurz gefasst
Fachartikel in der Zeitschrift PeerJ Computer Science, online erschienen am 05.06.2026. Verfasst von Wei Guan und 1 weiteren Autor:innen. Die Arbeit nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Hubei Business Service Development Research Center, Humanities and Social Sciences in Universities in Hubei Province.
Abstract (Originalsprache)
This article proposes a structured sparse graph learning method (SSGL) for unsupervised feature extraction, aimed at addressing three major challenges in high-dimensional economic data: difficulty in extracting variable interaction relationships, insufficient utilization of structural features, and severe noise interference. By integrating graph learning theory with sparse structural constraints, SSGL effectively captures complex correlations between economic variables and improves the robustness of feature representation. Firstly, by constructing an optimization objective function that integrates Frobenius norm reconstruction error, L1 sparse constraint, and kernel norm structural constraint, the Huber loss function is used to enhance noise robustness. Secondly, the SSGL method is used to construct the nearest neighbor graph between samples using representations to maintain the local structure of the data, and the least squares regression is used to model the global correlation structure of the data. Therefore, SSGL can simultaneously maintain the local and global correlation structures of data. In addition, SSGL uses sparse regularization to break the connections between different clustering samples in the graph, making the learned projections more discriminative. Finally, experiments on the Bureau of Economic Analysis (BEA) quarterly macroeconomic dataset showed that SSGL significantly outperformed seven mainstream baseline methods in clustering accuracy (88.8%), normalized mutual information (0.827), and adjusted Rand index (0.824). The reconstruction error was reduced by 36.2% compared to the optimal baseline, verifying its effectiveness and superiority in high-dimensional economic data analysis.
Förderung
- Hubei Business Service Development Research Center, Humanities and Social Sciences in Universities in Hubei Province
Zitieren
APA
Guan, W., & Yu, Y. (2026). A sparse and structured constraint-based method for large-scale economic data analysis. PeerJ Computer Science, 12, e3895. https://doi.org/10.7717/peerj-cs.3895
Vancouver
Guan W, Yu Y. A sparse and structured constraint-based method for large-scale economic data analysis. PeerJ Computer Science. 2026;12:e3895. doi:10.7717/peerj-cs.3895
BibTeX
@article{Guan2026,
author = {Guan, Wei and Yu, Yan},
title = {A sparse and structured constraint-based method for large-scale economic data analysis},
journal = {PeerJ Computer Science},
year = {2026},
volume = {12},
pages = {e3895},
doi = {10.7717/peerj-cs.3895}
}
EUVIMED